A Phase I/II Trial of Carfilzomib, Pegylated Liposomal Doxorubicin, and Dexamethasone for the Treatment of Relapsed/Refractory Multiple Myeloma

The past decade has seen a tremendous growth of available treatment options for relapsed refractory multiple myeloma (RRMM). Carfilzomib was approved by the FDA in 2012 for patients who had received at least two prior therapies including bortezomib and an immunomodulatory agent, based on results of a single-arm multicenter trial of 266 patients (1). Since that time, numerous clinical trials of novel combination regimens including carfilzomib have been conducted. One notable combination that the literature is currently lacking data on is carfilzomib and pegylated liposomal doxorubicin (PLD).

PLD is FDA approved in combination with bortezomib for RRMM. In the phase III randomized trial that was used to support its indication, PLD was shown to increase progression-free survival (PFS) by approximately 50% over bortezomib alone, but did not show an overall survival advantage in long-term follow-up (2, 3). Carfilzomib is a second-generation proteasome inhibitor (PI) that is an irreversible inhibitor of the 20S proteasome, which is structurally and mechanistically different from bortezomib. It is more selective for the chymotrypsin-like protease, causing less inhibitory activity against other active subunits. The ENDEAVOR trial randomized patients with RRMM to bortezomib and dexamethasone, or carfilzomib and dexamethasone, and demonstrated the superiority of carfilzomib over bortezomib with a median PFS with the carfilzomib regimen at 18.7 versus 9.4 months (4).

Given the additive or synergistic effect observed with PLD and bortezomib, we hypothesized that the combination of PLD and carfilzomib may be efficacious. To test this, we performed a prospective clinical trial using carfilzomib, PLD, and dexamethasone for RRMM. The study included both a dose-escalation phase to determine the MTD of the regimen, and an expansion phase powered to determine efficacy. In addition, we tested patient samples collected prior to treatment using a method called BH3 (Bcl-2 homology domain-3) profiling to identify predictive biomarkers for response.

A single-institution, open-label phase I/II clinical trial of carfilzomib, PLD, and dexamethasone was performed. Carfilzomib and dexamethasone were administered on days 1, 2, 8, 9, 15, and 16 and PLD was administered on day 8 of 28-day cycles. Up to 6 cycles of KDD induction treatment were administered. Subsequently, patients continued weekly carfilzomib and dexamethasone maintenance until disease progression or unacceptable toxicity.

The study was performed according to a protocol approved by the Washington University School of Medicine Human Subjects institutional review board (IRB). All subjects provided voluntarily written informed consent for the trial. The study was conducted in accordance with U.S. Common Rule. The study was registered at ClinicalTrials.gov as NCT01246063.

Eligible patients were 18 years of age or older with a diagnosis of RRMM post one or more lines of prior treatment and measurable disease in the blood (≥0.5 g/dL by electrophoresis or ≥10 mg/dL difference in involved and uninvolved light-chains), urine (≥200 mg/dL/24 hours), bone marrow (≥30% plasma cells), or extramedullary plasmacytoma. Patients with poor hematologic reserve, liver function, or performance status, and those requiring hemodialysis were excluded, as were patients with a history of plasma cell leukemia, HIV, or active hepatitis. Patients with a left ventricular ejection fraction below 55%, electrocardiogram evidence of acute ischemia or significant conduction system abnormalities, or recent history of acute myocardial infarction, unstable angina, or arrhythmia were excluded because of the known cardiac toxicities of carfilzomib and PLD. Patients who had received prior carfilzomib, PLD or doxorubicin, and those refractory to bortezomib were not excluded.

During the phase I portion of the trial, patients were enrolled in dose-escalating cohorts using a standard 3 + 3 design. Five dose levels were tested with carfilzomib doses ranging from 27 mg/m² to 56 mg/m². In all cohorts, the cycle 1 day 1 and 2 doses of carfilzomib were administered at a lead-in dose of 20 mg/m² and then escalated to the cohort specified level for all subsequent doses. All patients received 30 mg/m² PLD. The four initial cohorts did not receive dexamethasone to determine the MTD of carfilzomib-PLD doublet-therapy. The fifth cohort received 20 mg of dexamethasone in addition to carfilzomib and PLD at the previously determined MTD. The doses and schedule utilized are detailed in Fig. 1.

For MTD determination, dose-limiting toxicity (DLT) was defined as any of the following treatment emergent events related to study treatment during cycle 1 of induction treatment: grade 4 neutropenia or thrombocytopenia, febrile neutropenia, grade 2 neuropathy with associated pain; grade 3 or higher nonhematologic toxicity; or any other toxicity requiring dose reduction during cycle 1 or precluding cycle 2 of treatment on schedule.

On the basis of a priori sample size calculations, we determined that 24 patients treated at the MTD would allow us to determine whether the overall response rate (ORR, defined as partial response or better by standard International Myeloma Working Group Criteria) for patients treated with KDD was ≥40%, compared with the null hypothesis that ORR was ≤20%, at 0.1 alpha and 80% power. Therefore, additional patients were enrolled in the phase II expansion cohort until 24 patients total were treated at the MTD.

Bone marrow aspirate samples were collected prior to treatment and following discontinuation when possible. We assessed the apoptotic potential of the CD138⁺ plasma cells using mitochondrial profiling, or BH3 profiling, as a biomarker for identifying patients most likely to respond to study treatment.

Bone marrow samples were viably frozen at the time of collection. During the analytical profiling, the samples were thawed, Ficoll purified, and plasma cells were isolated using anti-CD138–coupled magnetic beads (Miltenyi Biotec). BH3 profiling was carried out in an ex vivo laboratory developed test by exposing derived plasma samples to peptides comprising the BH3-binding domains of Bcl-2 family proteins and BH3 mimetic compounds. The plasma cells were plated (∼10,000 cells/plate) and exposed to varying peptide derivatives from the BH3 family (BIM 100 μmol/L, BIM 0.1 μmol/L, PUMA 10 μmol/L, NOXA 100 μmol/L, BAD 100 μmol/L, HRK 100 μmol/L, BID 0.1 μmol/L, or MS-1 50 μmol/L) and with controls DMSO (1%) or carbonyl cyanide m-chlorophenyl hydrazone (10 μmol/L) upon permeabilization with digitonin and oligomycin. The mitochondria of treated cells were then stained with a fluorescent potentiometric mitochondrial membrane dye, JC-1 (Enzo Life Sciences), to measure the permeabilization of the mitochondrial outer membrane, which is the key signaling event in the apoptosis cascade. The change in relative fluorescence units of the membrane dye compared with the negative and positive controls was continuously recorded using a Tecan Infinite plate reader over a time course. The AUC of these readouts was determined. The results of each peptide were calculated as percent priming for death, indicating the dependence in a specific signaling pathway. Pearson correlations were used to compare priming of the Bcl-2 family proteins and BH3 mimetic compounds and Student t test was used to compare KDD responders and nonresponders.

Forty patients were enrolled, 23 in phase I and 17 in phase II, from 2012 through 2016. The study was considered complete in November 2017 and the remaining patients on trial at that time were transitioned to additional anti-multiple myeloma therapy at the discretion of their treating physician.

The baseline characteristics are summarized in Table 1. The median age of the study population was 65 years (range 270–79) and 58% were female. The median number of lines of prior therapy was 2.5 (range 1–13), with all patients receiving prior lenalidomide, 90% bortezomib, 10% carfilzomib, 8% PLD or doxorubicin, and 88% underwent a prior autologous stem cell transplant. Forty-five percent were refractory to bortezomib, 65% to lenalidomide, and 33% were refractory to both.

No DLTs were observed in the first four cohorts tested: (i) 27 mg/m² carfilzomib and 30 mg/m² PLD, (ii) 36 mg/m² carfilzomib and 30 mg/m² PLD, (iii) 45 mg/m² carfilzomib and 30 mg/m² PLD, (iv) 56 mg/m² carfilzomib and 30 mg/m² PLD. Thus, a MTD of doublet-therapy was not reached. Four patients were enrolled into cohorts 2 and 3 as 1 patient from each was removed from study for disease progression prior to completing the DLT evaluation period. Five patients were enrolled into cohort 4, and as there were no DLTs, it was determined that enrollment would continue with the fifth cohort.

The fifth cohort of patients received KDD triplet-therapy consisting of 56 mg/m² carfilzomib, 30 mg/m² PLD, and 20 mg of dexamethasone. Seven patients were enrolled, including one who was removed from the study based on disease progression prior to completing the DLT evaluation period. One of the six evaluable patients experienced grade 4 thrombocytopenia and was considered a DLT. On the basis of this, the MTD of KDD triplet-therapy was determined to be 56 mg/m² carfilzomib, 30 mg/m² PLD, and 20 mg of dexamethasone.

Mild to moderate hematologic toxicity was common (Fig. 2A); however, severe occurrences were infrequent and were generally treated adequately with growth factor support or transfusion and only resulted in dose reductions in 1 patient (3%). Grade 3/4 hematologic toxicity included: lymphopenia (63%), thrombocytopenia (40%), anemia (40%), neutropenia (28%), hemolysis (10%), and thrombotic thrombocytopenic purpura (TTP; 3%).

Gastrointestinal upset and constitutional symptoms were common but largely mild (Fig. 2B). The rate of palmar–plantar erythrodysesthesia syndrome was 15%. This led to PLD treatment discontinuation in 5%, and PLD dose reductions in another 5%. The incidence of thromboembolic events was 18% and there was one case of reversible posterior encephalopathy syndrome (RPLE). The only common grade 3/4 nonhematologic toxicity was infection (45%; Fig. 2C). Infectious events contributed to the two deaths on study. One patient died of sepsis and one died of acute respiratory failure secondary to H1N1 pneumonia. Both events occurred during cycle 3 of therapy. These events were not considered directly related to study treatment; however, the treatment may have contributed to their immunocompromised state.

As expected with carfilzomib and PLD administration, there were cardiopulmonary toxicities that included the following: dyspnea (55%), hypertension (33%), hypotension (18%), new or worsened congestive heart failure (5%), and myocardial infarction (3%; Fig. 2D). These events led to the discontinuation of study treatment in 8% of patients. These events occurred following the DLT observation period in later cycles of treatment.

Overall, 73% (29/40) of patients treated on the study had a confirmed response, including 40% with a VGPR or better. For determining efficacy, the analysis was limited to patients treated at the MTD. At the MTD, the ORR was 83% (20/24), including 54% (13/24) with a very good partial response (VGPR) or better, and 25% (6/24) obtaining a complete response (CR/stringent CR). The median number of cycles administered was 9.5 (range 1–34). Eleven patients were removed from study due to disease progression. The estimated median PFS, defined as time to progression, was 13.4 months (95% CI, 5.0–21.7). Patients who began alternative anti-multiple myeloma treatment prior to progression were censored. The estimated median overall survival was not reached after a median follow-up of 23.3 months. Efficacy data are summarized in Table 2 and the survival curves are depicted in Fig. 3.

Twenty-two patients had received prior bortezomib, 10 were refractory, and 12 were sensitive. KDD was effective in patients with bortezomib-refractory disease with 60% (6/10) having an objective response. All bortezomib-responsive/naïve patients responded (4 CR/sCR, 6 VGPR, and 4 PR). In bortezomib and lenalidomide (double)-refractory patients, the ORR was 50% (3/6). KDD was also effective among patients with high-risk features. The ORR for patients with high-risk disease by ISS (stage 3) or mSMART criteria was 77% (10/13).

Thirty-two patient samples, 28 pretreatment and 4 posttreatment, met the QC requirements and underwent BH3 profiling. Unless otherwise specified, all data and analyses were limited to the pretreatment samples. There was a varying degree of correlation between the readout of the assay, “priming” of the various Bcl-2 family proteins determined by exposure to various BH3 mimetic compounds (r² = 0.183–0.986). Clinical response (partial response or better) was associated with lower priming for three of the analytes. Median NOXA priming was 15.4% among responders compared with 26.2% for nonresponders (P = 0.0383), PUMA priming 30.5% compared with 43.3% (P = 0.013), and HRK priming 20.8% compared with 59.6% (P = 0.001; Fig. 4A). Moreover, higher HRK priming was associated with inferior PFS; for each 10-point increase the risk of progression increased by 37% (HR 1.39; 95% CI, 1.09–1.72; P = 0.007). Those with HRK priming in the highest quintile (>40.0%) had a median estimated PFS of 1.8 months compared with 12.4 months for all other patients (P < 0.001; Fig. 4B). Three patients had paired pre- and posttreatment samples; all three initially responded to treatment, but later discontinued because of disease progression. There was a trend for increased HRK priming following discontinuation as compared with pretreatment with a median of 30.5% pretreatment and 41.8% posttreatment (P = 0.0374; Fig. 4C). HRK priming and all other Bcl-2 family proteins and BH3 mimetic compounds tested were similar between bortezomib-refractory and bortezomib-sensitive/naïve patients (Fig. 4D).

This phase I/II trial of KDD is the first report of a triplet regimen with high-dose carfilzomib (56 mg/m²) in patients with RRMM. KDD appeared well tolerated and efficacious in RRMM. The estimated ORR of the regimen is 83% (95% CI, 68%–98%) with a median PFS of 13.4 months (95% CI, 5.0–21.7). The treatment appeared active across high-risk subgroups of patients such as those refractory of bortezomib or with high-risk features.

Although it is difficult to compare between trials, it is important to interpret results in the context of other carfilzomib-containing salvage regimens. The ORR of 83% seen with KDD is comparable with other studies of high-dose carfilzomib including an ORR of 77% with K_d in the ENDEAVOR trial (4, 5). However, the ORR in subjects that were not refractory to bortezomib was 100% in this study suggesting the addition of PLD improves response compared with K_d alone. KDD also appeared to elicit deeper responses with a CR/sCR rate of 25% compared with 13% in the ENDEAVOR trial. The ENDEAVOR trial showed a median PFS of 18.7 months compared with 13.4 months with KDD. Of note, the study reported here enrolled subjects with a median of 2 prior lines of therapy and 42% were refractory to bortezomib, compared to a median of 1 prior line of therapy and only 3% bortezomib refractory in the ENDEAVOR trial. This may account for part of the differences in PFS. In addition, the ENDEAVOR trial continued carfilzomib on the traditional day 1, 2, 8, 9, 15, and 16 schedule until disease progression. In this study of KDD, the carfilzomib schedule was reduced to once weekly after cycle 6 for convenience. PLD was also limited to just six cycles based on the potential for cumulative cardiotoxicity from PLD in combination with carfilzomib. The durability of responses in this study may have been improved had carfilzomib/PLD been continued on the induction schedule.

Another popular carfilzomib triple regimen, carfilzomib (27 mg/m²), lenalidomide, and dexamethasone (KRd), showed an ORR of 87% with a PFS of 26.3 months in a population with a median of 2 prior therapies and where 15% of patients were refractory to bortezomib and 6% were double-refractory to bortezomib and lenalidomide (6). In comparison, the bortezomib and double-refractory rates were 42% and 25%, respectively, in the KDD trial. In the KRd trial, carfilzomib was continued on the traditional day 1, 2, 8, 9, 15, and 16 schedule for 12 cycles. The differences in prior treatments, the additional doses of carfilzomib, and the prolonged exposure to lenalidomide, which was continued until progression, make comparisons difficult.

In addition, a comparison of this regimen to prior reports of PLD and bortezomib combination is relevant (2). In this prior study, two-thirds of subjects had been treated with ≥2 prior lines of therapy and all subjects were bortezomib naïve. The ORR was 44% (4% CR and 40% PR) and median PFS was 9 months. The side-effect profile of PLD and bortezomib appears similar to KDD but response rates are higher with KDD.

The combination of KDD appeared well tolerated. Grade 3/4 hematologic toxicity included neutropenia (28%), thrombocytopenia (40%), and anemia (40%). This compares with the combination of PLD and bortezomib, which reported similar rates of neutropenia (29%), thrombocytopenia (23%), and anemia (9%) (2). The combination of high-dose carfilzomib and dexamethasone alone reported grade 3/4 thrombocytopenia (8%) and anemia (14%) (4). There were, however, cases of hemolysis, TTP, and RPLE suggesting potential for endothelial injury with KDD.

The exact mechanism for endothelial injury is unknown, but others have speculated it could be related to direct effects of proteasome inhibition on NF-κB; or, by causing impairment of vasodilation as well as oxidative and inflammatory stress; or, through an immune-mediated mechanism (7, 8). We believe this effect is independent of PLD and not potentiated by the combination. When the safety of the current study is compared in the context of the ENDEAVOR trial, the addition of PLD to carfilzomib increased hematologic toxicity, but showed similar incidence and types of nonhematologic toxicity compared to K_d alone (9, 10). This toxicity profile suggests this regimen should be reserved for patients with adequate bone marrow reserve.

Cardiac adverse events are a concern particular to carfilzomib relative to bortezomib with a review of single-agent carfilzomib studies showing a 7.2% incidence of cardiac failure (11). Trials with high-dose carfilzomib reported similar rates of cardiac failure (4, 12). Combining carfilzomib with an anthracycline may cause concern for an increase in cardiac toxicity, but this was not observed with KDD with only 5% (2/40) of patients having any grade of new or worsening cardiac failure.

A potential next step in the evolution of multiple myeloma treatment is better selection of the available therapies based on biomarkers predictive of response to treatment. In this study, we analyzed the use of mitochondrial profiling as a prognostic biomarker for response of KDD. The underlying principle of the assay is that aberrant phenotypes in cancer cells lead to dependence on certain Bcl-2 proteins for survival. The assay identifies which protein is involved in cell survival by measuring the ability of various BH3 mimetic proteins to induce apoptosis. These interactions occur primarily through BH3-mediated binding. This indirectly determines the predisposition of that cell to respond to drug-induced apoptosis signals and is called “mitochondrial priming” (13).

Measurements of the mitochondrial priming state have been found to associate with patient response to treatments in multiple myeloma, AML, DLBCL, and CLL (14–20). These associations have been seen for a range of chemotherapies and targeted drugs including regimens that target antiapoptotic proteins. The efficacy of the Bcl-2–selective BH3 mimetic compound venetoclax, for instance, is linked to the extent of Bcl-2 dependence (19). The efficacy of the Mcl-1–targeted BH3 mimetics, on the other hand, is linked to Mcl-1 dependence (20). These dependencies can impact any treatment that ultimately relies on mitochondrial apoptosis for efficacy.

Here, we have seen that Bcl-xL dependence (high HRK priming), identified in the BH3 profiling assay, was associated with poorer response and inferior outcomes following KDD treatment. It is currently unclear whether this is a possible mechanism for resistance to KDD, as detailed in Fig. 5, or a nonspecific indicator of more aggressive disease biology. There was a trend of increased priming following disease progression on KDD in the limited number of patients, where paired samples were available. However, BH3 profiling was similar in bortezomib-refractory and sensitive/naïve patients prior to KDD. This finding has been previously reported in bortezomib-refractory/naive U266 cell lines (21). This may suggest that the Bcl-xL dependence is specific to KDD and not a class effect of PIs. Moreover, knowing the HRK score may help screen patients for alternative treatment with Bcl-xL–targeted therapies in future studies. The use of BH3 profiling after treatment with such regimens as KDD might also be used to identify survival dependencies and guide the use of Mcl-1–targeted therapies or Bcl-2–targeted therapy like venetoclax (22, 23).

The strength of this study is its prospective design. The limitations of this study include its nonrandomized design and small sample size, which make additional analysis such as multivariate infeasible. The original sample size calculation for this study was based on a null hypothesis of <20% ORR, which we concede may be an underestimate given recent studies of mAb combinations and small-molecule inhibitors showing evidence of higher activity in patients with dual bortezomib and lenalidomide-refractory myeloma, but was based on contemporary data at time of protocol design. Despite the limitations, the data suggest that KDD is a viable salvage option beyond second-line for patients with multiple myeloma who are PI exposed or refractory with adequate bone marrow reserve. As mAb regimens such as daratumumab + pomalidomide and dexamethasone, are increasingly used in second- and third-line before carfilzomib-based regimens, the KDD regimen could provide an alternative to continuing pomalidomide in a successive line of therapy.

In conclusion, in this trial a three-drug regimen of KDD with carfilzomib administered at 56 mg/m² was efficacious and well tolerated in patients with RRMM. Biomarkers such as BH3 profiling may help determine which patients are most likely to benefit from the regimen; however, additional study is needed to validate this assay.

M.A. Schroeder reports receiving speakers bureau honoraria from Merck and Takeda, and is a consultant/advisory board member for Amgen, Incyte, Sanofi/Genzyme, Pfizer and Partner Therapeutics. M.H. Cardone has ownership interests (including patents) in Eutropics Pharmaceuticals, and reports receiving commercial research grants from NCI-SBIR. T. Wildes reports receiving commercial research support from Janssen and provided consultation to Carevive. K.E. Stockerl-Goldstein reports receiving speakers bureau honoraria from Celgene, and is a consultant/advisory board member for Janssen. R. Vij is a consultant/advisory board member for Amgen. No potential conflicts of interest were disclosed by the other authors.

Conception and design: M.A. Fiala, R. Vij

Development of methodology: M.A. Fiala, S.R. Jean, R. Vij

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M.A. Schroeder, M.A. Fiala, S.R. Jean, K. Shea, A. Ghobadi, T. Wildes, K.E. Stockerl-Goldstein, R. Vij

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M.A. Schroeder, M.A. Fiala, E. Huselton, M.H. Cardone, S. Jaeger, S.R. Jean, T. Wildes, R. Vij

Writing, review, and/or revision of the manuscript: M.A. Schroeder, M.A. Fiala, E. Huselton, M.H. Cardone, S.R. Jean, A. Ghobadi, T. Wildes, K.E. Stockerl-Goldstein, R. Vij

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M.A. Schroeder, M.A. Fiala, K. Shea

Study supervision: M.A. Fiala, R. Vij

We would like to thank the Alvin J. Siteman Cancer Center at Washington University School of Medicine and Barnes-Jewish Hospital in St. Louis, MO, for the use of the Clinical Trials Core that provided protocol development services. We would also like to thank the members of Washington University School of Medicine Division of Oncology Clinical Trials office who provided study coordination and data management. The Siteman Cancer Center is supported, in part, by an NCI Cancer Center Support Grant #P30 CA91842. BH3 profiling was supported by NCI-SBIR R44CA203610 awarded to Eutropics. Amgen Inc. provided funding and materials for this study.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.